An Overview of Atmospheric
Analyses and Reanalyses
for Climate
Kevin E. Trenberth
NCAR
Boulder CO
Analysis
Data Assimilation merges observations & model predictions
to provide a superior state estimate.
x
t
 dynamics physics x
It provides a dynamically- consistent estimate of the state
of the system using the best blend of past, current, and
perhaps future observations.
Obs
4DDA
Model
Improved
products,
predictions,
understanding
Experience mainly in atmosphere; developing in ocean, land
surface, sea ice.
Data assimilation system
 The observations are used to correct errors in the
short forecast from the previous analysis time.
 Every 12 hours ECMWF assimilates 7 – 9,000,000
observations to correct the 80,000,000 variables that
define the model’s virtual atmosphere.
 This is done by a careful 4-dimensional interpolation in
space and time of the available observations; this
operation takes as much computer power as the 10-day
forecast.
ECMWF 2009
NWP models and data assimilation continues to improve
Operational forecast scores of major NWP centers. RMSE of
geopotential height at 500hPa in NH (m) for 24-hour forecasts
are displayed. The scores of forecasts have improved over time.
NWP Forecast skill scores continue to improve
Extratropical NH and SH forecasts: 12 month means
plotted at last month. Updated from Simmons and Hollingsworth 2002
SH skill became comparable to NH after about 2002!
Reanalysis
Operational four dimensional data assimilation
continually changes as methods and assimilating models
improve, creating huge discontinuities in the implied
climate record.
Reanalysis is the retrospective analysis onto global
grids using a multivariate physically consistent
approach with a constant analysis system.
Reanalysis has been applied to atmospheric data covering the past
five decades. Although the resulting products have proven very
useful, considerable effort is needed to ensure that reanalysis
products are suitable for climate monitoring applications.
From: Executive Summary of “The Second Report on the Adequacy of The
Global Observing Systems for Climate in Support of the UNFCCC”.
20 Aug 2003
Atmospheric Reanalyses
Current atmospheric reanalyses, with the horizontal resolution
(latitude; T159 is equivalent to about 0.8 ), the starting and ending
dates, the approximate vintage of the model and analysis system, and
current status.
Reanalysis
Horiz.Res
Dates
Vintage
Status
NCEP/NCAR R1
T62
1948-present
1995
ongoing
NCEP-DOE R2
T62
1979-present
2001
ongoing
CFSR (NCEP)
T382
1979-present
2009
thru 2009, ongoing
C20r (NOAA)
T62
1875-2008
2009
Complete, in progress
ERA-40
T159
1957-2002
2004
done
ERA-Interim
T255
1989-present
2009
ongoing
JRA-25
T106
1979-present
2006
ongoing
JRA-55
T319
1958-2012
2009
underway
1979-present
2009
thru 2010, ongoing
MERRA (NASA) 0.5
8
What have we gained and what are the benefits?
Prior to reanalyses, the analyzed climate record was beset
with major discontinuities from changes in the data
assimilation systems. It was difficult, if not impossible, to
reliably infer anomalies and to analyze climate variability.
The use of a stable data assimilation system has produced
fairly reliable records for monitoring, research and
improved prediction that have enabled :
 climatologies to be established
 anomalies to be reliably established
 time series, empirical studies and quantitative diagnostics
 exploration of, improved understanding of processes
 model initialization and validation
 test bed for model improvement on all time scales,
especially seasonal-to-interannual forecasts
 Greatly improved basic observations and data bases.
What have we learned?
Observing system changes affect variability
Trends and low frequencies unreliable
Exacerbated by model bias
Budgets don’t balance
Impacts many diagnostic studies
Problems with hydrological cycle
Sensitivity to model physics (e.g., convection)
Exacerbated by insertion of observations
Problems with warm season continental climates
precipitation
diurnal cycle
Unrealistic surface fluxes
Ocean (radiative, freshwater)
Land (precipitation, radiative)
Limits usefulness for offline forcing; e.g. ocean modeling
Limits ability to do coupled assimilation
Quantities/regions not a priority for weather centers
Surface
Stratosphere
Polar regions
Many aspects of tropics
Reanalysis
A MAJOR challenge remains the
continually changing observing system in
spite of substantial improvements in
bias correction in the latest generation
of reanalyses
11
Satellite Data Streams assimilated
Dec
TOVS
Feb
Jul
TIROS-N
Apr
Sep
Nov Oct
Feb
NOAA-6
NOAA-7
NOAA-8
May Jun Jul Oct
Jan
Nov
NOAA-9
NOAA-10
Dec
Sep
Nov
Jan
Sep Sep
Sep
NOAA-11
NOAA-12
Jun
NOAA-14
Jan
Dec
NOAA-15 Sep
ATOVS
NOAA-16 Nov
NOAA-17 Jul
NOAA-18
EOS Aqua
EOS Aqua
Oct
GOES-08
GOES
Sounders
Apr
Jul
GOES-10
Apr
Jun
F08
GOES-12
Jul
SSM/I
Nov
F10
Dec
Dec
F11
Jul
Nov
Dec
Dec
F13
May
F14
May
F15 Dec
Aug
78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08
Example: Satellite based observations
•
•
•
•
•
•
Satellites typically last 3-5 years and have to be replaced
Orbits decay
Equator crossing times change
New satellite orbits differ
Instrument calibrations drift and can be changed by launch
Interference can occur from other instruments
The Changing Observing System
1973
77k/6h
1987
550k
1979– 77K Obs
324k
1973
every
6hrs
2006
4,220k
1987
– 550K Obs every
6hrs
1979 – 325K Obs every 6hrs
2006 – 4.2M Obs every 6hrs
Bias corrections are needed
But how good are they?
Is there a baseline to establish real trends?
Bias corrections should be applied to satellite
and radiosonde data.
Potential for unintended perturbations or bad data
to be perpetuated.
Most radiosonde stations do NOT have adequate records
of changes
Need to document bias correction changes to almost
all observing systems.
Bias correction procedures have greatly improved
Top: Global mean bias estimates for MSU channel 2 computed in ERA-Interim using
new bias correction procedures (top) and recorded warm-target temperatures used
for on-board instrument calibration (bottom) show remarkable agreement
Dee et al 2009.
Examples of results from reanalyses
with emphasis on problems
Surface Temperature: filled in gaps
Ten year mean anomalies in 2 m temperature (K) relative to the 1989–
1998 mean for (a) CRUTEM3 for 1979–1988, (b) ERA-40 for 1979–
1988, (c) CRUTEM3 for 1999–2008, and (d) ERA Interim for 1999–
2008. Reanalysis values are plotted for all 5 grid squares for which
there are CRUTEM3 data and for all other grid squares with more
than 10% land.
Simmons et al 2010.
Missing data for CRUTEM3 => underestimate trends vs full ERA
Focus on:
MERRA and
ERA-I
Which have
smooth
evolving
moisture
fields
(no spinup):
•4Dvar
•nudging
Precipitable water
Precipitation errors in reanalyses
Bosilovich et al 2008
MERRA
Identifiable discons:
•SSM/I mid-1987,
•TOVS to ATOVS:
AMSU-A,AMSU-B
late 1998 to 2001
(NOAA 15 =>NOAA 12
NOAA 16 => NOAA 14, March
20, 2001),
AIRS late 2002,
GPS RO 2002 on,
COSMIC April 2006.
Precipitation
Freshwater flux E-P
From moisture budget
Transport
E-Pocean
P-Eland
28
Energy budget:
Reanalyses
ASR bias 1990s
Biggest in summer
 All reanalyses have too
much incoming solar radiation
in southern oceans
 Caused by too few clouds
 Implies too much heating of
ocean which should diminish
poleward heat transports
when models are coupled
 Has implications for storm
tracks and ocean transports
Trenberth and Fasullo 2010
0.6
30
Energy budget: Reanalyses
 At TOA, most climate models are tuned to get
balance or replicate ERBE/CERES
 Depends on equilibrium simulation
 No longer works in reanalyses
 Specified SSTs
 Global imbalances (hide even bigger local)
Resolution
ASR
OLR
Net(TOA)
Net (sfc)
R1 ERA-40 ERA-I JRA MERRA CFSR
1.9° 0.8° 0.5° 1.1°
0.5°
0.5°
-13
-2
-12
-3
-1
+6
-8
+4
+5
+6
-2
+6
+8
+16
-8
-8
+6
+3
+2
+13
+5 W m-2
+4
0
+8
For 1990s vs climatology
31
Reanalyses
 Even if the assimilating model has a balanced
energy budget, when SSTs are specified
there is an infinite heat and moisture source
or sink
 There is no feedback on the SSTs from
surface fluxes
 The result is potentially large energy
imbalances at TOA and at surface
 The TOA and surface energy balances can
be strong diagnostics of model bias problems
Reanalysis
1. The next (4th) International reanalysis
conference is planned to be in April 2012 in
Washington DC area.
2. There is not a problem with lack of reanalyses,
indeed there is a proliferation. The problems
are:
1. lack of an end to end program with adequate
evaluation of products (and the funding), and
2. Reanalysis is all done in a research domain and not
sustained, so that key personnel can be lost.
3. Lack of adequate vetting and diagnosis
3. Reanalysis is an essential part of climate
services, especially in monitoring, attribution
and prediction
33
World Climate Research Programme